Estimating housing consumption adjustments from panel data

Although time-series analysis of panel data has generally not been used in the study of housing demand, longitudinal analysis provides the opportunity to increase our understanding of housing adjustments over time. In particular, time series analysis of housing choice using panel data permits appropriate use of both time-series and cross-sectional variation in housing prices, provides the opportunity to separate age and cohort effects, and facilitates integration of mobility analysis with housing choice analysis. In this paper, a longitudinal discrete choice model of the choice of housing tenure and size is presented and estimated using panel data. This conditional fixed effects multinomial logit model, developed by G. Chamberlain, is computationally convenient and successfully accounts for time-invariant differences among households. While household-specific unobserved characteristics can be readily accounted for in a linear model, this is not the case for a quantitative choice model. The use of a model of the type employed here thus provides a crucial link between time series analysis and a discrete choice setting. Estimation of this model yields results with respect to age and price that differ from results obtained from individual and pooled cross-sections. This provides support for the plausible proposition that housing choices are intertemporally correlated, and, more importantly, emphasizes the importance of accounting for this correlation to estimate consistently the parameters of housing choice models.